There is an ever greater need to forecast correlations in addition to standard deviations, both for managing risk exposure in a portfolio context such as value at risk, and for pricing derivatives whose payoffs depend on correlations among more than one random factor. Like volatilities, correlations can be estimated from historical security returns data or as implied values from market prices of options. In this article, Walter and Lopez consider using implied volatilities from sets of three foreign exchange options to compute the correlations that they imply. They examine the implied correlations that can be obtained from two currency trios, dollar-deutschemark-Swiss franc, and dollar-deutschemark-Japanese yen, and compare them against correlations estimated from standard statistical models applied to historical data, including GARCH. There is considerable time variation in measured correlations, but none of the estimation methods clearly dominates the others. Implied correlations appear to be more useful for the trio including the yen than for the one including the Swiss franc.
Stored only when both papers are in the corpus — 77 citations exist upstream.